On Mon, 8 Feb 2016 19:15:43 +0100, Marcos Cruz wrote:

> Some interesting BogoMIPS benchmarks from the Spanish QL Forum
> (http://foro.speccy.org/viewtopic.php?f=15&t=4687):
> 
> QL with GoldCard (1):
> 1.62, 1.62, 1.62, ...
> 
> QL with SuperGoldCard (2) :
> 5.83, 5.83, 5.83, ...
> 
> QemuLator (3):
> 96.55, 97.25, 96.55, 97.61, 97.08
> 
> QPC2 (3):
> 123.56, 123.98, 123.56, 123.56, 123.70, 123.27
> 
> SMSQmulator (3):
> 52.32, 49.98, 49.43, 50.45, 49.43
> 
> Notes:
> (1) GoldCard -> Motorola 68000 16 MHz.
> (2) SuperGoldCard -> Motorola 68020 24 MHz.
> (3) PC - Intel Core2 Quad 2.33 GHz - Windows 10.

More benchmarks, with even more exotic machines, for the fun of it:

QPC II v4.02/Linux 4.1/Wine 1.7.42 (*):
270.46 BogoMips
121.458 VAX Mips/213401.6 Dhrystones/s (10 millions runs)

QPC II v4.02/VirtualBox 5.0.14/Windows XP Pro SP3 (*):
267.09 BogoMips
120.481 VAX Mips/211685.0 Dhrystones/s (10M runs)

SMSQmulator v3.15/SUN Java 7 (JRE 1.7.0.80) (*):
131.26 BogoMips
54.579 VAX Mips/95895.7 Dhrystones/s (5M runs)

Q60 @ 66MHz (overclocked 68060RC50):
127.82 BogoMips
Writethrough cache mode: 24.937 VAX Mips/43813.5 Dhrystones/s (5M runs)
Copyback cache mode: 47.812 VAX Mips/84005.4 Dhrystones/s (5M runs)

QXL @ 35MHz (overclocked 68040RC33):
21.47 BogoMips
13.169 VAX Mips/23137.4 Dhrystones/s (1M runs)

(*) PC using an overclocked Core-i5 2500K, perma-locked in turbo mode
(i.e. only the C0, C1 and P0 CPU/Package states are allowed: frequency
stepping is fully disabled) at 4.6GHz, running under Linux v4.1
(32 bits + PAE mode).

All tests were done without job running under SMSQ/E, with SMSQ/E v3.21
for all but QPC II (v3.19) and SMSQmulator (3.23), all at a resolution
of 1024x768 (but for the Q60, which resolution is 1024x512).

Note that the Dhrystone/VAX Mips figures come from GCCdhrystone v2.1:
they are more relevant of the actual power you can get from a machine
since BogoMips only represents the maximum instructions per second
throughput of a CPU executing the simplest instructions in its set
(increment-test-branch loop). However, the performance reported by
Dhrytstone v2.1 depends on what compiler optimizations the C compiler
that was used to compile it is capable of. The cross-compiled GCC
version (GCCdhrystone) is therefore (much) faster than the C68 one
(dhrystone).

Thierry.
_______________________________________________
QL-Users Mailing List

Reply via email to